International Comparison CCQM K46 – Ammonia in Nitrogen · 2010-08-23 · International...

International Comparison CCQM K46 – Ammonia in Nitrogen Adriaan M.H. van der Veen1, Gerard Nieuwenkamp1, Rob M. Wessel1, Masaaki Maruyama2, Gwi Suk Heo3, Yong-doo Kim3, Dong Min Moon3, Bernhard Niederhauser4, Manuela Quintilii4, Martin J.T. Milton5, Maurice G. Cox5, Peter M. Harris5, Franklin R. Guenther6, George C. Rhoderick6, L.A. Konopelko7, Y.A. Kustikov7, V.V. Pankratov7, D.N. Selukov7, V.A. Petrov7, E.V. Gromova7 1Van Swinden Laboratorium (VSL), Thijsseweg 11, 2629 JA Delft, the Netherlands 2Chemicals Evaluation and Research Institute, Japan (CERI), 1600, Shimo-Takano, Sugito-machi, Kitakatsushika-gun, Saitama 345-0043, Japan

3Korea Research Institute of Standards and Science (KRISS), Division of Metrology for Quality Life, P.O.Box 102, Yusong, Taejon, Republic of Korea 4Bundesamt für Metrologie (METAS), Lindenweg 50, 3003 Bern-Wabern,Switzerland 5National Physical Laboratory (NPL), Teddington, Middlesex, TW11 0LW, United Kingdom 6 National Institute of Standards and Technology (NIST), Gas Metrology Group 100 Bureau Drive Building 301, Gaithersburg, MD 20899-3574, United States of America 7D .I. Mendeleyev Institute for Metrology (VNIIM), Department of State Standards in the field of Physical Chemical Measurements, 19, Moskovsky Prospekt, 198005 St-Petersburg, Russia

Field Amount of substance

Subject Ammonia in nitrogen

Table of contents Field ....................................................................................................................................................1 Subject ................................................................................................................................................1 Table of contents.................................................................................................................................1 Introduction.........................................................................................................................................2 Participants..........................................................................................................................................2 Measurement standards.......................................................................................................................2 Measurement protocol ........................................................................................................................2 Schedule..............................................................................................................................................2 Mixture preparation ............................................................................................................................3 Validation measurements....................................................................................................................3 Measurement methods ........................................................................................................................5 Results.................................................................................................................................................6 Model for the key comparison ............................................................................................................7 References.........................................................................................................................................13 Annex A: Justification of procedure for evaluating this key comparison.........................................15 Annex B: Measurement report NPL .................................................................................................17 Annex C: Measurement report KRISS..............................................................................................22 Annex D: Measurement report CERI................................................................................................34 Annex E: Measurement report METAS............................................................................................37 Annex F: Measurement report VNIIM .............................................................................................40 Annex G: Measurement report VSL .................................................................................................44 Annex H: Measurement report NIST................................................................................................46

Introduction Ammonia is an important compound in chemical industry. It is widely used and is the basis for producing other compounds containing nitrogen. Ammonia is also very hazardous, and consequently emissions of ammonia need be controlled and monitored. In the past years, several National Metrology Institutes have developed facilities for the preparation of Primary Standard gas Mixtures (PSMs), dynamically generated ammonia mixtures and facilities for comparing and certifying gas mixtures containing ammonia.

This report describes the results of a key comparison for ammonia in nitrogen. The amount–of–substance fraction level of ammonia chosen for this key comparison is 30-50 µmol/mol. This key comparison aims to support CMC-claims for ammonia from 30 µmol/mol onwards.

Participants Table 1 lists the participants in this key comparison.

Table 1: List of participants

Acronym Country Institute CERI JP Chemicals Evaluation and Research Institute, Saitama, Japan KRISS KR Korea Research Institute of Standards and Science, Daejeon,

Republic of Korea METAS CH Federal Office of Metrology, Wabern, Switzerland VSL NL Van Swinden Laboratorium B.V., Delft, the Netherlands NPL UK National Physical Laboratory, Teddington, Middlesex, United

Kingdom NIST US National Institute of Standards and Technology, Gaithersburg

MD, United States of America VNIIM RU D.I. Mendeleyev Institute for Metrology, St. Petersburg,

Russia

Measurement standards A set of mixtures was prepared gravimetrically by VSL. The mixtures were verified against a set of VSL PSMs and in addition validated by a dynamic method.

The pressure in the cylinders was approximately 80 bar; cylinders of 5 dm3 nominal were used. The amount-of-substance fractions as obtained from gravimetry and purity verification of the parent gases were supposed to be used as reference values. In due course of the key comparison, it turned out that the assigned amount fractions could not be regarded as the best realisation of SI, so that a consensus-based approach was developed, taking into consideration the small differences in the amount fractions as obtained from the preparation data.

The nominal amount-of-substance fraction was 30-50 µmol/mol.

Measurement protocol The measurement protocol requested each laboratory to perform at least 3 measurements, with independent calibrations. The replicates, leading to a measurement, were to be carried out under repeatability conditions. The protocol informed the participants about the nominal concentration ranges. The laboratories were also requested to submit a summary of their uncertainty evaluation used for estimating the uncertainty of their result.

Schedule The schedule of this key comparison was as follows:

February 2006 Draft protocol to participants October 2006 Registration of participants October 2006 Preparation of gravimetric mixtures + first verification measurement November 2006 Second verification December 2006 Third verification January 2007 Shipment of cylinders to participating laboratories May 15, 2007 Reports due to pilot laboratory May 31, 2007 Cylinders due to pilot laboratory June 2007 Final verification measurement

Mixture preparation The intention was to use as reference values the ones based on gravimetry, and the purity verification of the parent gases. All mixtures underwent verification prior to shipping them to the participants. After return of the cylinders, they have been verified once more to reconfirm the stability of the mixtures.

In the preparation, the following four groups of uncertainty components have been considered:

1. gravimetric preparation (weighing process) (xi,grav)

2. purity of the parent gases (Δxi,purity)

3. stability of the gas mixture (Δxi,stab)

4. correction due to partial recovery of a component (Δxi,nr)

The amount of substance fraction xi,prep of a particular component in mixture i, as it appears during use of the cylinder, can now be expressed as

,,,,,, nristabipurityigraviprepi xxxxx ΔΔΔ +++= (1)

The value obtained from equation (1) is sometimes referred to as “gravimetric value”. Assuming independence of the terms in equation (1), the expression for the combined standard uncertainty becomes

2, nristabipurityigraviprepi uuuuu +++= . (2)

For the mixtures used in this key comparison, the following statements hold (for all components involved). First of all, the preparation method has been designed in such a way that

,0, =Δ nrix (3)

and its standard uncertainty as well. Furthermore, long-term stability study data has shown that

,0, =Δ stabix (4)

and its standard uncertainty is estimated from the validation measurements.

Validation measurements After preparation the mixtures have been validated four times over a period of four months to evaluate the mixture stability. After return the mixtures have again been analysed to control the stability. Mixtures were analysed by a calibrated photo-acoustic analyzer. Calibration was performed in the range from 30-40 µmol/mol using 2 PSMs of 30 µmol/mol and 2 PSMs of 40 µmol/mol. In Table 2 the results of the 4 measurements are summarized. In figure 1, the results are visualized calculated as recovery being the results of analysis divided by the gravimetric value.

Figure 2 shows the mean value of the 4th validation measurements.

Table 2: Validation results (µmol mol-1)

cylinder number

Preparation

October analysis

November analysis

December analysis

January analysis

D751992 33.929 33.9 33.8 33.9 33.6 D752114 33.983 34.2 34.2 34.2 34.1 D751916 34.028 34.2 34.1 34.0 34.0 D751986 34.057 34.0 33.8 33.6 33.6 D751946 34.091 34.0 33.8 33.7 33.9 D751929 34.104 34.2 34.1 33.7 34.2 D751936 34.113 34.0 34.1 33.6 34.0

analysis compared to gravimetric mixtures

0 2 4 6 8

cylinder number

y Series1

Series2

Series3

Series4

Figure 1: Results of 4 validation measurements

Mean values validation analyses

0 2 4 6 8

Cylinder number

Figure 2: Mean values of 4 validation measurements

As an independent extra validation step, a dynamic preparation method was set-up to validate the gravimetric values of the 10 mixtures. For this dynamic method (ISO 6145-4 [13]) a gas tight syringe was filled with a 50 mmol/mol NH3 in N2 mixture and injected in a calibrated flow of nitrogen. The injection was controlled by a motor driven injector. The injection volume per time was calibrated with a stable gas component (carbon monoxide). By repeated filling of the syringe 6 mixtures were

prepared in the calibrated range of the photo-acoustic analyser. The results of the 4th analysis are given in figure 3.

Results 4th NH3 analysis

0 2 4 6 8 10 12 14

Cylinder / syringe number

cylinder

syringe

Figure 3: Results of 4th analysis including 6 dynamically prepared mixtures

After return of the cylinders a final analysis was performed to check the stability. The results were compared to the initial validation analysis and visualised in figure 4.

Stability CCQM-K46 NH3 mixtures

NIST VNIIM VSL METAS Ceri NPL Kriss

participant

beforeafter

Figure 4: Results before and after shipment of the cylinders

Measurement methods The measurement methods used by the participants are described in annex A of this report. A summary of the calibration methods, dates of measurement and reporting, and the way in which metrological traceability is established is given in table 3.

Table 3: Summary of calibration methods and metrological traceability

Laboratory Calibration Traceability Measurement technique CERI ISO 6143 3 CERI PSMs ND-IR KRISS Single point KRISS CRM ND-IR METAS ISO 6143 Permeation system; dilution of

KC mixture to 550 nmol/mol Photo acoustic IR

VSL ISO 6143 9 VSL PSMs Photo acoustic IR NPL Bracketing 2 NPL PSMs Photo acoustic IR NIST ISO 6143 Permeation system; multiple

permeation tubes Chemiluminescence

VNIIM OLS1 Own standards UV absorption

Results This comparison involves N = 7 participant laboratories using three methods of measurement, static gravimetry (method 1), dynamic gravimetry (method 2) and static gravimetry with pre-bleaching (method 3). The laboratories that apply method j are termed laboratory group j.

The results are given in table 4.

Table 4: Results submitted in this key comparison

Mixture Laboratory xiprep

µmol mol-1 u(xi

prep) µmol mol-1

µmol mol-1 U(xi

lab) µmol mol-1

Method

D751916 VSL 34.028 0.201 33.8 0.7 2 1 D751929 NPL 34.104 0.201 34.52 0.99 2 1 D752114 VNIIM 33.983 0.201 33.7 0.5 2 1 D751986 METAS 34.057 0.201 32.27 0.36 2 2 D751992 NIST 33.929 0.201 32.37 0.28 2 2 D751936 KRISS 34.113 0.201 32.91 0.62 2 3 D751946 CERI 34.091 0.201 32.81 0.38 2 3

where xi

prep amount of substance fraction, from gravimetric preparation ui

prep standard uncertainty associated with xiprep accounting for stability

(uistab = 0.2 µmol mol-1)

xilab measured value for laboratory i

U(xilab) expanded uncertainty associated with xi

1 Ordinary least squares

Result KCRV method 1 method 2 method 3

x ilab -x

Laboratory

Figure 5: Results and key comparison reference value

The data can be specified as follows

( )( )ii xux , , i = 1, …, N, where preplabiii xxx −= and ( ) ( ) ( )prep2lab22

iii xuxuxu +=

The standard uncertainty associated with xilab is obtained in the usual way by dividing the expanded

uncertainty by the coverage factor k.

The data are shown in figure 5. On the y-axis, the differences with respect to the assigned amount–of–substance fraction on the basis of mixture preparation are given.

Model for the key comparison Each xi relating to a particular method estimates the sum of the KCRV and a method-related bias quantity. There are four parameters in the model: the quantity of which the KCRV is a best estimate and three method bias quantities.

There are only three independent parameters since an arbitrary constant can be added to the KCRV and subtracted from the bias quantities without essentially changing the solution. This lack of independence can be resolved by introducing a constraint. Here the sum of the biases is chosen to be zero.

The value component of the ith degree–of–equivalence is taken as xi minus the calculated KCRV. Correlations are taken into account when evaluating the uncertainty components of the degrees–of–equivalence.

The procedure can be summarised as follows. Given are measured values xi and associated standard uncertainties u(xi), i = 1, …, 7, and the knowledge that laboratories in group 1 (i = 1, 2, 3), group 2 (i = 4, 5) and group 3 (i = 6, 7) have different biases.

1. Form weights

( )ii xuw 2−= , i = 1, …, 7.

2. Calculate weighted means of the three groups:

( ) ( )3213322111 wwwxwxwxwv ++++= ,

( ) ( )5455442 wwxwxwv ++= ,

( ) ( )7677663 wwxwxwv ++= ,

and evaluate the associated standard uncertainties u(v1), u(v2), u(v3) using

( ) 321121 wwwvu ++= ,

( ) 54221 wwvu += ,

( ) 76321 wwvu += .

3. Test the consistency of the measured values with respect to the relevant weighted means [14].

4. Calculate the (overall) KCRV and evaluate the associated standard uncertainty u(xref) using

( ) 3321ref vvvx ++= ,

( ) ( ) ( ) ( )[ ] 932

ref2 vuvuvuxu ++= .

5. Calculate biases for the three groups:

ref11 xvs −= , ref22 xvs −= , ref33 xvs −= ,

and evaluate the associated standard uncertainties u(s1), u(s2), u(s3) using

( ) ( ) ( ) ( )[ ] 94 32

12 vuvuvusu ++= ,

( ) ( ) ( ) ( )[ ] 94 32

22 vuvuvusu ++= ,

( ) ( ) ( ) ( )[ ] 94 32

12 vuvuvusu ++= .

6. Form the degrees–of–equivalence (di, 2u(di)), i = 1, …, 7, using

refxxd ii −= , ( ) ( ) ( ) ( ) ( )[ ] 95 32

1222 vuvuvuxudu ii ++−+= ,

for i = 1, 2, 3,

refxxd ii −= , ( ) ( ) ( ) ( ) ( )[ ] 95 32

1222 vuvuvuxudu ii +−++= ,

for i = 4, 5, and

refxxd ii −= , ( ) ( ) ( ) ( ) ( )[ ] 95 32

1222 vuvuvuxudu ii −++= ,

for i = 6, 7.

The results from applying the above procedure are given below. The data and weights are given in table 5. Table 5: Key comparison data and assigned weights

Mixture Lab x µmol mol-1

u(x) µmol mol-1

w µmol-2 mol2

Method

D751916 VSL -0.228 0.40 6.14 1 D751929 NPL 0.416 0.53 3.50 1 D752114 VNIIM -0.283 0.32 9.72 1 D751986 METAS -1.787 0.27 13.74 2 D751992 NIST -1.559 0.24 16.68 2 D751936 KRISS -1.203 0.37 7.33 3 D751946 CERI -1.281 0.28 13.08 3

The method means (vj) and biases (sj) are given in table 6.

Table 6: Method means and biases

Method vj µmol mol-1

u(vj) µmol mol-1

sj µmol mol-1

u(sj) µmol mol-1

1 -0.14 0.23 0.88 0.18 1.33 2 -1.66 0.18 -0.64 0.16 0.39 3 -1.25 0.22 -0.23 0.18 0.03

The values for χ2 are in all cases smaller than the critical values with a probability of 0.05 and 2 (method 1) respectively 1 degree of freedom (methods 2 and 3). The results are consistent with the method means.

The reference value is –1.018 µmol mol-1. The associated standard uncertainty is 0.122 µmol mol-1. The χ2 value for the method means is 28.7 and exceeds by far the critical value with a probability of 0.05 and 2 degrees of freedom (χ2

crit = 5.99).

The degrees–of–equivalence are given in table 7 and shown in figure 6.

Table 7: Degrees–of–equivalence

Mixture Lab di µmol mol-1

u(di) µmol mol-1

U(di) µmol mol-1

D751916 VSL 0.790 0.38 0.757 D751929 NPL 1.434 0.52 1.031 D752114 VNIIM 0.735 0.29 0.577 D751986 METAS -0.769 0.26 0.513 D751992 NIST -0.541 0.23 0.460 D751936 KRISS -0.185 0.34 0.689 D751946 CERI -0.263 0.24 0.484

IIM VS

Laboratory

Figure 6: Degrees–of–equivalence relative to the KCRV

Discussion

The evaluation of the data in this key comparison had to be revisited. In the protocol of this key comparison, it was envisaged to use the data from static gravimetric preparation as key comparison reference value (KCRV), which had been successful in a preceding comparison between NPL, VNIIM and VSL. The experimental work undertaken prior to releasing the gas mixtures for this key comparison, including the independent validation through the syringe injection method, seemed to confirm the amount fractions assigned by the procedures described in ISO 6142 [7]. In retrospect, it can only be concluded that these experiments were not sufficiently accurate to take away all doubts concerning the validity of the preparation data.

The stratification of the data along the lines of the three different methods and subsequent data evaluation highlights that the three methods used provide different results. Within a stratum, the data submitted are nicely consistent, but between strata the differences are substantial.

The biggest differences are observed between the static gravimetric preparation with and without pre-bleaching of the cylinder. In a supplementary report [10], the passivation process is described to be carried out with 500 µmol mol–1 ammonia in nitrogen. After evacuation with a pressure below 5 Pa, the cylinders are used for preparing mixtures of ammonia in the range 100 µmol mol–1 down to 20 µmol mol–1. The quantity of residual gas is so small, that it cannot explain the difference between the results of method 1 (gravimetric preparation without pre-bleaching) and method 3.

Adsorption effects of ammonia can have a more profound effect on the amount fraction of gas as sampled from the cylinder. These effects impact method 1, where an apparent loss can occur as well as method 3, where ammonia adsorbed to the wall can be released. Additional work by CERI [11] confirms the difference between methods 1 and 3 (table 8). It is important to note that the cylinders without treatment are not of the same type as the ones used in this key comparison.

Table 8: Supplementary results static gravimetric preparation [11]

Pre-treatment

xiprep

µmol mol-1 xi

µmol mol-1 time 0 days 3 days 3 weeks CPB-18498 none 34.85 32.04 30.9 30.68

-8.06% -11.33% -11.97% CPB-18499 none 34.85 30.28 28.75 28.41

-13.11% -17.50% -18.48% xi

prep xianal xi

anal xianal

time 2 days 1 week 3 weeks CPB-18498 passivated 35.26 35.08 34.95 34.99

-0.51% -0.88% -0.77% CPB-18499 passivated 34.89 34.7 34.53 34.59

-0.54% -1.03% -0.86%

The results of CERI do not provide independent proof of its validity; the PSMs used for calibration are likely to be affected in the same manner as the mixtures prepared in the passivated cylinders. The losses observed in the untreated cylinders are larger than the discrepancy between methods 1 and 3 in this key comparison, which may be due to the different cylinder wall treatments performed by the cylinder manufacturers.

Further work by the coordinating laboratory has demonstrated equivalence, within the respective uncertainties, of method 1 with the continuous injection method (ISO 6145-4 [13]) [12]. The implementation of the syringe method was validated using carbon monoxide (figures 7 and 8) [12].

80 µmol/mol CO in N2

dynamic vs gravimetry

0 2 4 6 8 10 12

syringe

gravimetry

Figure 7: Comparison of syringe injection with gravimetry at 80 µmol/mol CO in N2 mixtures [12]

0 2 4 6 8 10 12

syringe

gravimetry

Figure 8: Comparison of syringe injection with gravimetry at 40 µmol/mol CO in N2 mixtures [12]

The results show the validity of the results of the syringe method with an expanded uncertainty of 1% relative. The comparisons between gravimetry and the syringe method for ammonia are shown in figures 9 and 10.

300 µmol/mol NH3 in N2

0 2 4 6 8 10 12

syringe

gravimetry '01

gravimetry '05

Figure 9: Comparison of syringe injection with method 1 at 300 µmol/mol NH3 in N2 [12]

0 2 4 6 8 10 12 14

syringe

gravimetry '01

gravimetry '05

Figure 10: Comparison of syringe injection with method 1 at 30 µmol/mol NH3 in N2 [12]

The setup of the validation experiments precludes that effects of, e.g., wall adsorption are fully propagated. Therefore, it can be said that method 1 is supported by the continuous injection method (ISO 6145-4 [13]). This conclusion was the basis for the proposal of CCQM-K46 in the first place.

After the first presentation of the results of this key comparison, NIST stated that the cylinder has a side connection (DIN-1) that is different than their usual size. This means that the pressure regulator used for this comparison was not of the preferred type. VSL has sent one of their regulators to NIST.

With this pressure regulator NIST observed a smaller discrepancy (–3% relative instead of –4.6% relative, see table 4) with respect to the value from preparation for the key comparison mixture.

The results of the syringe method and method 1 are however not consistent with methods 2 and 3. All four methods are designed in a way that they should be able to deliver metrologically traceable results. This key comparison demonstrates that notwithstanding the strong belief about the validity of the results and methods with the participants, there are serious discrepancies in the measurement data. This discrepancy is clearly reflected in the value of χ2 computed at the level of the KCRV. The value of χ2 exceeds by far the critical value at the 5% level. Hence, the consistency criterion as defined for a weighted mean is violated [14].

This key comparison aims to support CMC claims for ammonia in nitrogen or air in the range from 30 µmol/mol to 100 µmol/mol.

Conclusions

The results in this key comparison do not show true consensus. Three different methods have been used, which give results that are discrepant. When grouped in accordance with the employed methods, the data are consistent. This key comparison cannot resolve the observed differences between the laboratories. Further experimental work is needed.

As KCRV the mean of the three methods is used. In its uncertainty, no allowance is made for the observed biases. With respect to the KCRV, only two laboratories report consistent results: CERI and KRISS.

References [1] Alink A., “The first key comparison on Primary Standard gas Mixtures”, Metrologia 37

(2000), pp. 35-49

[2] Van der Veen A.M.H., Brinkmann F.N.C., Arnautovic M., Besley L., Heine H.-J., Lopez Esteban T., Sega M., Kato K., Kim J.S., Perez Castorena A., Rakowska A., Milton M.J.T., Guenther F.R., Francy R., Dlugokencky E., “International comparison CCQM-P41 Greenhouse gases. 1. Measurement capability”, Metrologia 44 (2007), Techn. Suppl. 08002

[3] BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, OIML, “Guide to the expression of uncertainty in measurement”, first edition, ISO Geneva, 1995

[4] CIPM, “Mutual recognition of national measurement standards and of calibration and measurement certificates issued by national metrology institutes”, Sèvres (F), October 1999

[5] BIPM, Annex B to the MRA, http://kcdb.bipm.fr/BIPM-KCDB//AppendixB/

[6] Alink A., Van der Veen A.M.H., “Uncertainty calculations for the preparation of primary gas mixtures. 1. Gravimetry”, Metrologia 37 (2000), pp 641-650

[7] International Organization for Standardization, ISO 6142:2001 Gas analysis - Preparation of calibration gas mixtures - Gravimetric methods, 2nd edition

[8] Van der Veen A.M.H., Pauwels J., “Uncertainty calculations in the certification of reference materials. 1. Principles of analysis of variance”, Accreditation and Quality Assurance 5 (2000), pp. 464-469

[9] Van der Veen A.M.H., Linsinger T.P.J., Lamberty A., Pauwels J., “Uncertainty calculations in the certification of reference materials. 3. Stability study”, Accreditation and Quality Assurance 6 (2001), pp. 257-263

[10] Maruyama M., Uehara S., “Passivation processes for ammonia gas standards in CERI”, CERI, December 2008, filed as GAWG09/13

[11] Maruyama M., Uehara S., “An additional study of NH3 gas standards”, CERI, December 2008, filed as GAWG09/14

[12] Nieuwenkamp G., Wessel R.M., “CCQM K46 – Ammonia in nitrogen. Overview of the results and additional actions”, filed as GAWG09/47

[13] International Organization for Standardization, “ISO 6145-4:2004 Gas analysis - Preparation of calibration gas mixtures using dynamic volumetric methods - Part 4: Continuous syringe injection method”, ISO, Geneva, 2004

[14] Cox M.G., “The evaluation of key comparison data”, Metrologia 39 (2002), pp. 589–595

Coordinator Van Swinden Laboratorium Department of Chemistry Rob M. Wessel Thijsseweg 11 2629 JA Delft the Netherlands Phone +31 15 2691 677 Fax +31 15 261 29 71 E-mail RWessel@vsl.nl

Project reference CCQM-K46

Completion date October 2007

Annex A: Justification of procedure for evaluating this key comparison The model is given by

1ref SXX i += , i = 1, 2, 3 (Group 1),

2ref SXX i += , i = 4, 5 (Group 2),

3ref SXX i += , i = 6, 7 (Group 3),

Xi quantity for which xi is the measured value and u(xi) the associated standard uncertainty provided by laboratory i,

Xref quantity for which the KCRV is the best estimate, S1, S2, S3 bias quantities for methods 1, 2 and 3. Reference value(s) and bias quantities Make the changes of variables

1ref1 SXV += , 2ref2 SXV += , 3ref3 SXV += . (5)

The model becomes

1VXi = , (Group 1),

2VX i = , (Group 2),

3VX i = , (Group 3). Because each of these three sub-models has the same structure as a model for a key comparison in which each laboratory measures a single, stable artefact, each Vk can be regarded as a reference value for the kth group of laboratories.

A best estimate vk of Vk is the weighted mean of the values xi in group k. The standard uncertainty u(vk) associated with vk can be established as described in reference [14]. As the groups are mutually independent there is no correlation associated with the vk.

Summing expressions (5),

321ref321 3 SSSXVVV +++=++ . (6)

As a result of the resolving constraint, which can be written as

0321 =++ SSS ,

expression (6) gives

( ) 3321ref VVVX ++= . (7)

Hence the KCRV xref and its associated standard uncertainty u(xref) are given by

( ) 3321ref vvvx ++= , ( ) ( ) ( ) ( )[ ] 932

ref2 vuvuvuxu ++= .

Bias quantities

Using expressions (5) and (7), the bias quantity for method 1, with similar expressions for methods 2 and 3, is

( ) 32 321ref11 VVVXVS −−=−= ,

and so the method biases s1, s2 and s3 and the associated standard uncertainties u2(s1), u2(s2) and u2(s3) are given by

ref11 xvs −= , ( ) ( ) ( ) ( )[ ] 94 32

12 vuvuvusu ++= ,

ref22 xvs −= , ( ) ( ) ( ) ( )[ ] 94 32

22 vuvuvusu ++= ,

ref33 xvs −= , ( ) ( ) ( ) ( )[ ] 94 32

12 vuvuvusu ++= .

Degrees of equivalence

Take the degrees–of–equivalence with respect to the overall KCRV:

refXXD ii −= .

Accordingly, the degrees–of–equivalence can be expressed as

( )( )ii dud 2, , refxxd ii −= , ( ) ( )ref22 xxudu ii −= . (8)

Now, using expression (7),

( ) 3321ref VVVXXX ii ++−=− . (9)

The quantities on the right-hand side of expression (9) are mutually independent apart from Xi and Vk, where k = 1, 2 or 3 as appropriate. Accordingly, for group 1,

( ) ( ) ( ) ( ) 993 32

ref2 vuvuvxuxxu ii ++−=− . (10)

The first term on the right-hand side of expression (10) can be written as u2(xi – θv1) with θ = 1/3. When θ = 1, this term is identical in form to that in expression (8), used in evaluating the uncertainty component of the degree–of–equivalence for laboratory i. By a similar analysis to that in appendix C of reference [1], a modest generalization of that expression is

( ) ( ) ( ) ( )122

12 2 vuxuvxu ii θθθ −−=− .

For the case in hand (θ = 1/3),

( ) ( ) ( ) 953 122

12 vuxuvxu ii −=− .

So, applying this result to expression (10) and extending to all groups,

( ) ( ) ( ) ( ) ( ) 9995 32

1222 vuvuvuxudu ii ++−= (Group 1),

( ) ( ) ( ) ( ) ( ) 9959 32

1222 vuvuvuxudu ii +−+= (Group 2),

( ) ( ) ( ) ( ) ( ) 9599 32

1222 vuvuvuxudu ii −++= (Group 3).

Annex B: Measurement report NPL Report from the National Physical Laboratory CCQM-K46 (Ammonia in nitrogen)

Reference Method The CCQM-K46 standard (D751929) was compared with two NPL gravimetric standard using an Innova 1301 Gas Analyser based on photoacoustic absorption.

Sample Handling Cylinders were connected directly to a gas sampling system (with no regulators) and purged prior to connection to the gas analyser via 1/8” 316L stainless steel tubing. The configuration of the sampling system and analyser are shown below:

Innova 1301

1/8” stainless steel tubing

Manual 3-way ball switching

NPL standard

standard

Needle valve

Atmospheric bypass vent

(approximately 2.5 l/min)

Data Processing Comparisons were made by completing the following sequence:

A1-B1-A2-B2-A3-B3-A4

Where ‘An’ is the measurement of the amount fraction of NH3 in the NPL standard and ‘Bn’ is the measurement of the amount fraction of NH3 in the CCQM-K46 standard and ‘n’ is the number of repeats of each measurement made of an individual standard. The result of each of these measurements (An and Bn) was derived from the average of four individual measurements. A value for the amount fraction of NH3 in the CCQM-K46 standard was then calculated as follows:

1 1∑

Where Ar

n is the result from measurement An of the amount fraction of NH3 in the NPL standard and Br

n is the result from measurement Bn of the amount fraction of NH3 in the CCQM-K46 standard and x is the reference amount fraction of NH3 in the NPL standard (from gravimetry). NPL Standards The two standards used for this inter-comparison are listed below. In each case Air Products VLSI NH3 was used. The principal impurities specified by the manufacturer are (in μmol/mol): N2(3), H20 (2), CO (1), CO2 (1), O2 (0.5) and Ar (0.5).

NPL No 30636 Component Supplier Grade Amount fraction

(μmol/mol) NH3 Air Products VLSI 34.37 N2 Air Products BIP+ Balance

The standard was prepared in a 10 litre aluminium cylinder supplied by Scott Gases, which had been treated using the Aculife III process. The standard was produced by the dilution of an appropriate quantity of a standard (#30588), which had itself been prepared from a standard at approximately 5 % prepared using “traditional” cascade dilution techniques.

30588 x U (x )'Ar' 4.991 14.401'CO' 0.027 0.010'CO2' 0.027 0.010'H2O' 0.024 0.010'CxHy' 0.050 0.030

'N2' 997925.822 14.495'CH4' 0.076 0.009'O2' 0.001 0.000

'NH3' 2000.172 1.879

NPL No 30575 Component Supplier Grade Amount fraction

(μmol/mol) NH3 Air Products VLSI 39.98 N2 Air Products BIP+ Balance

The standard was prepared in a 10-litre aluminium cylinder supplied by Scott Gases, which had been treated using the Aculife III process. The standard was produced using a single step “loop injection” method (of 38.02g +/- 0.1g). The cylinder was rolled for 1 hour after preparation and prior to analysis in order to ensure the homogeneity of the standard.

The single-step loop injection method used to prepare this standard was a departure from the traditional cascade dilution methods that have been employed at NPL and were used previously for the preparation of sub 100 μmol/mol NH3 standards. NPL 30575 was compared with NPL 30577 (39.97 μmol/mol), a standard prepared using traditional dilution methods, in order to validate the new approach. The two standards were compared and found to agree within the estimated uncertainties of their gravimetric values (see section entitled ‘Results’).

Uncertainty in the Reference Values The dominant uncertainty in the reference values of the gravimetric NH3 standards is that due to loss of material to the walls of the cylinder. Our best estimate of this effect imposes a measurement uncertainty of 1% (relative).

'N2' 999885.888 14.429'CH4' 0.075 0.009'O2' 0.000 0.000

'NH3' 39.975 0.105

'N2' 999891.499 14.184'CH4' 0.075 0.009'O2' 0.000 0.000

'NH3' 34.363 0.060

Results Validation of single-step preparative method (NPL 30577 vs NPL 30575)

Results of analyses NPL 30575 vs D751929(CCQM)

Results of analyses NPL 30636 vs D751929(CCQM)

Summary of results

Measurement Measured mole fraction

[μmol/mol]

1 34.16 2 34.52 3 34.35 4 34.99 5 34.49 6 34.62

Mean 34.52 Standard deviation 0.28

Estimation of Measurement Uncertainty

Source of uncertainty

Analytical repeatability 2

1⎟⎟⎠

⎞⎜⎜⎝

⎛⋅

Non-linearity of analyser 0.22 Uncertainty in reference values due to gravimetry and stability 0.32

Standard uncertainty (k=1) 49.03.02.0628.0 22

=++⎟⎟⎠

⎞⎜⎜⎝

Expanded uncertainty (k=2) 0.99

Final result for D751929

34.52 μmol/mol ± 0.99 μmol/mol (k=2)

Annex C: Measurement report KRISS

Report Form CCQM-K46 Ammonia in nitrogen Laboratory name: KRISS (written by Gwi Suk Heo)

Cylinder number: D751936

Initial inner pressure of Cylinder as received : 1100 PSIG Date of reception : March 1, 2007 Name of the contact person : Gwi Suk Heo (e-mail address : heo@kriss.re.kr)

Measurement #1 Component Date

(dd/mm/yy) Result (umol/mol)

Standard deviation (% relative)

number of replicates

Ammonia 04/05/07 32.92 0.01 10

Ammonia 04/05/07 32.91 0.03 10

Ammonia 04/05/07 32.92 0.01 10

Results Component Result

(umol/mol) Expanded Uncertainty

(umol/mol)

Coverage factor

Ammonia 32.91 0.62 2.0

KRISS NDIR raw data in analysis of K-46 sample *conc in excel below not corrected for adsorption loss of NH3. The adsorption loss was corrected and

recalculated during the uncertainty evaluation of the K-46 sample analysis.

Analy.data 070514-1KRISS CRM K-46 cyl

Cyl. No D518955 D751936 D518955 D751936 D518955File No 70001 70001 80001 80001 90001

onc.(m mol/m 34.154 33.241 34.154 33.245 34.154

1 33.87 32.93 33.86 32.94 33.822 33.87 32.93 33.86 32.94 33.823 33.87 32.93 33.84 32.94 33.824 33.87 32.99 33.84 32.92 33.815 33.87 32.99 33.84 32.92 33.81

Mean area 33.870 32.954 33.848 32.932 33.816u 0.9917 0.99104 0.9901

S.T.D 0.00 0.03 0.01 0.01 0.01R.S.D(%) 0.00 0.10 0.03 0.03 0.02R.U(%) 0.00 0.07 0.02 0.02 0.01

Analy.data 070514-2Cyl. No D518955 D751936 D518955 D751936 D518955File No 60001 60001 70001 70001 80001

onc.(m mol/m 34.154 33.224 34.154 33.238 34.154

1 34.39 33.46 34.38 33.45 34.362 34.40 33.46 34.38 33.45 34.363 34.40 33.43 34.35 33.45 34.364 34.40 33.43 34.35 33.42 34.355 34.38 33.43 34.35 33.42 34.35

Mean area 34.394 33.442 34.362 33.438 34.356u 1.0070 1.0061 1.0059

S.T.D 0.01 0.02 0.02 0.02 0.01R.S.D(%) 0.03 0.05 0.05 0.05 0.02R.U(%) 0.02 0.03 0.03 0.03 0.01

Analy.data 070514-3Cyl. No D518955 D751936 D518955 D751936 D518955File No 50001 50001 60001 60001 70001

onc.(m mol/m 34.154 33.244 34.154 33.241 34.154

1 34.27 33.35 34.22 33.34 34.322 34.27 33.35 34.22 33.34 34.323 34.27 33.34 34.22 33.36 34.324 34.28 33.34 34.27 33.36 34.265 34.28 33.34 34.27 33.36 34.26

Mean area 34.274 33.344 34.240 33.352 34.296u 1.0035 1.0025 1.0042

S.T.D 0.01 0.01 0.03 0.01 0.03R.S.D(%) 0.02 0.02 0.08 0.03 0.10R.U(%) 0.01 0.01 0.06 0.02 0.07

Uncertainty evaluation for preparation of KRISS NH3 CRM and analysis of K-46 NH3 sample Model equation: CK46 = (Ca + Cb + Cc) / 3 ; Ca = ((Ra1 / ((Rax1+Rax2)/2) * xA) + (Ra2 / ((Rax2+Rax3)/2) * xA)) / 2 * fpress ; Cb = ((Rb1 / ((Rbx1+Rbx2)/2) * xA) + (Rb2 / ((Rbx2+Rbx3)/2) * xA)) / 2 * fpress ; Cc = ((Rc1 / ((Rcx1+Rcx2)/2) * xA) + (Rc2 / ((Rcx2+Rcx3)/2) * xA)) / 2 * fpress ; xA = (nA2 / nT) * 1000000 * (fads * fhomo2 * fstab2) ; nT = nA2 + nNofm + nN2 ; nA2 = (mA2 / MT2) * (xA1/100) ; nNofm = (mA2 / MT2) * (xN1/100) ; MT2 = MA * (xA1/100) + MN * (xN1/100) ; nN2 = mN2 / MN ; mA2 = (WbalA2 + δWug / 1000000 + δWhadug / 1000000) ; mN2 = (WbalN2 + δWg / 1000 + δWhad / 1000) ; xA1 = (nA1 / nT1 * 100) * fpurity * fhomo1 * fstab1 ; xN1 = nN1 / nT1 * 100 ; nT1 = nA1 + nN1 ; nA1 = mA1 / MA ; nN1 = mN1 / MN ; mA1 = (WbalA1 + δWg / 1000 + δWhad / 1000) ; mN1= (WbalN1 + δWg / 1000 + δWhad / 1000) ; List of quantities:

Quantity Unit Definition

CK46 ppm conc of K-46 from analysis (mean value) Ca ppm conc of K-46 from analysis Cb ppm conc of K-46 from analysis Cc ppm conc of K-46 from analysis Ra1 peak area of sample from analysis Rax1 peak area of std from analysis Rax2 peak area of std from analysis xA ppm conc of KRISS NH3 std gas Ra2 peak area of sample from analysis Rax3 peak area of std from analysis fpress factor for pressure difference in sample introduction Rb1 peak area of sample from analysis Rbx1 peak area of std from analysis Rbx2 peak area of std from analysis Rb2 peak area of sample from analysis Rbx3 peak area of std from analysis

Rc1 peak area of sample from analysis Rcx1 peak area of std from analysis Rcx2 peak area of std from analysis Rc2 peak area of sample from analysis Rcx3 peak area of std from analysis nA2 mole mole of NH3 in 30 ppm std gas (kriss) nT total mole of gases in 30 ppm NH3 std gas fads factor related to loss of NH3 by adsorption in 30 ppm std gas

fhomo2 uncertainty related to homogeneity of 30 ppm std gas fstab2 uncertainty related to stability of 30 ppm NH3 std gas nNofm mole mole of N2 contained 2% NH3 used in preparation of 30 ppm

NH3 nN2 mole mole of N2 used for preparation of 30 ppm NH3(as dilution gas) mA2 g wt(g) of 2% NH3 gas taken for prep of 30 ppm gas MT2 g/mol MW of 2% NH3 gas taken for prep of 30 ppm gas xA1 % conc of NH3 at 2% NH3 gas xN1 % conc of N2 at 2% NH3 gas MA g/mol MW of NH3 MN g/mol MW of N2 mN2 g wt of N2 dil gas in prep of 10 ppm gas

WbalA2 g wt of 2% NH3 std gas taken for prep of 30 ppm gas δWug ug uncertainty related to chemical balance measurement

δWhadug ug uncertainty related to handling of cylinder in balance measurement

WbalN2 g wt of N2 taken for prep of 30 ppm gas δWg mg uncertainty related to chemical balance measurement

δWhad mg uncertainty related to handling of cylinder in balance measurement

nA1 mole mole of NH3 in 2% NH3 std gas nT1 mole total mole of gases in 2% std gas

fpurity uncertainty of NH3 purity fhomo1 uncertainty related to homogeneity of 2 % std gas fstab1 uncertainty related to stability of 2% H2S std gas nN1 mole mole of N2 used in prep of 2% std gas mA1 g wt of NH3 in prep of 2% std gas mN1 g wt of N2 used in prep of 2% std gas

WbalA1 g wt of NH3 in prep of 2% std gas

WbalN1 g wt of N2 used in prep of 2% std gas CK46: Result Ca: Interim result Cb: Interim result Cc: Interim result Ra1: Type B t-distribution Value: 32.954 Standard uncertainty: 0.0147 Degrees of freedom: 4 Rax1: Type B t-distribution Value: 33.870 Standard uncertainty: 0.0000 Degrees of freedom: 4 Rax2: Type B t-distribution Value: 33.848 Standard uncertainty: 0.0049 Degrees of freedom: 4 xA: Interim result Ra2: Type B t-distribution Value: 32.932 Standard uncertainty: 0.0049 Degrees of freedom: 4 Rax3: Type B t-distribution Value: 33.816 Standard uncertainty: 0.0024 Degrees of freedom: 4 fpress: Type B normal distribution Value: 1

Expanded uncertainty: 0.002 Coverage factor: 2 Rb1: Type B t-distribution Value: 33.442 Standard uncertainty: 0.0073 Degrees of freedom: 4 Rbx1: Type B t-distribution Value: 34.394 Standard uncertainty: 0.0040 Degrees of freedom: 4 Rbx2: Type B t-distribution Value: 34.362 Standard uncertainty: 0.0073 Degrees of freedom: 4 Rb2: Type B t-distribution Value: 33.438 Standard uncertainty: 0.0073 Degrees of freedom: 4 Rbx3: Type B t-distribution Value: 34.356 Standard uncertainty: 0.0024 Degrees of freedom: 4 Rc1: Type B t-distribution Value: 33.344 Standard uncertainty: 0.0024 Degrees of freedom: 4 Rcx1: Type B t-distribution Value: 34.274 Standard uncertainty: 0.0024 Degrees of freedom: 4 Rcx2: Type B t-distribution Value: 34.240 Standard uncertainty: 0.0122 Degrees of freedom: 4 Rc2:

Type B t-distribution Value: 33.352 Standard uncertainty: 0.00489898 Degrees of freedom: 4 Rcx3: Type B t-distribution Value: 34.296 Standard uncertainty: 0.0147 Degrees of freedom: 4 nA2: Interim result nT: Interim result fads: Type B normal distribution Value: 1 Expanded uncertainty: 0.022 Coverage factor: 2 fhomo2: Type B normal distribution Value: 1 Expanded uncertainty: 0.0068 Coverage factor: 2 fstab2: Type B normal distribution Value: 1 Expanded uncertainty: 0.0002 Coverage factor: 2 nNofm: Interim result nN2: Interim result mA2: Interim result MT2: Interim result xA1: Interim result xN1: Interim result

MA: Type B normal distribution Value: 17.03056 g/mol Expanded uncertainty: 0.000255604 g/mol Coverage factor: 2 MN: Type B normal distribution Value: 28.01348 g/mol Expanded uncertainty: 0.000161658 g/mol Coverage factor: 2 mN2: Interim result WbalA2: Type B t-distribution Value: 1.04436 g Standard uncertainty: 0.000012603 g Degrees of freedom: 82 δWug: Type B normal distribution Value: 0 ug Expanded uncertainty: 100 ug Coverage factor: 2 δWhadug: Type B normal distribution Value: 0 ug Expanded uncertainty: 1000 ug Coverage factor: 2 WbalN2: Type B t-distribution Value: 616.33531 g Standard uncertainty: 0.00013254 g Degrees of freedom: 74 δWg: Type B normal distribution Value: 0 mg Expanded uncertainty: 20 mg Coverage factor: 2 δWhad: Type B normal distribution Value: 0 mg Expanded uncertainty: 10 mg Coverage factor: 2

nA1: Interim result nT1: Interim result fpurity: Type B normal distribution Value: 0.9997 Expanded uncertainty: 0.000133 Coverage factor: 2 fhomo1: Type B normal distribution Value: 1 Expanded uncertainty: 0.0018 Coverage factor: 2 fstab1: Type B normal distribution Value: 1 Expanded uncertainty: 0.002 Coverage factor: 2 nN1: Interim result mA1: Interim result mN1: Interim result WbalA1: Type B t-distribution Value: 8.0074 g Standard uncertainty: 0.000205 g Degrees of freedom: 91 WbalN1: Type B t-distribution Value: 644.0219 g Standard uncertainty: 0.00000023 g Degrees of freedom: 98 Uncertainty budget: red color uncertainty factors are major factors.

Quantity Value Standard uncertainty

Degrees of freedom

Sensitivity coefficient

Uncertainty contribution

Corr.-coeff.

Ca 32.917 ppm 0.309 ppm Cb 32.905 ppm 0.309 ppm

Degrees of freedom

Corr.-coeff.

Cc 32.916 ppm 0.309 ppm Ra1 32.9540 0.0147 4 0.166 2.45·10-3

ppm 0.01 0.000

Rax1 33.87 0.0 4 0.0 0.0 ppm 0.0 0.0 Rax2 33.84800 4.90·10-3 4 -0.162 -794·10-6

ppm 0.00 0.000

xA 33.819 ppm 0.316 ppm Ra2 32.93200 4.90·10-3 4 0.167 816·10-6

ppm 0.00 0.000

Rax3 33.81600 2.40·10-3 4 -0.0811 -195·10-6 ppm

0.00 0.000

fpress 1.00000 1.00·10-3 50 32.9 0.0329 ppm 0.11 0.011Rb1 33.44200 7.30·10-3 4 0.164 1.20·10-3

ppm 0.00 0.000

Rbx1 34.39400 4.00·10-3 4 -0.0797 -319·10-6 ppm

0.00 0.000

Rbx2 34.36200 7.30·10-3 4 -0.160 -1.16·10-3 ppm

0.00 0.000

Rb2 33.43800 7.30·10-3 4 0.164 1.20·10-3 ppm

0.00 0.000

Rbx3 34.35600 2.40·10-3 4 -0.0798 -192·10-6 ppm

0.00 0.000

Rc1 33.34400 2.40·10-3 4 0.165 395·10-6 ppm

0.00 0.000

Rcx1 34.27400 2.40·10-3 4 -0.0801 -192·10-6 ppm

0.00 0.000

Rcx2 34.2400 0.0122 4 -0.160 -1.95·10-3 ppm

-0.01 0.000

Rc2 33.35200 4.90·10-3 4 0.164 806·10-6 ppm

0.00 0.000

Rcx3 34.2960 0.0147 4 -0.0800 -1.18·10-3 ppm

0.00 0.000

nA2 752.86·10-6 mole

1.48·10-6 mole

nT 22.038959 405·10-6 fads 0.99000 5.00·10-3 50 33.2 0.166 ppm 0.54 0.289

fhomo2 1.00000 7.00·10-3 50 32.9 0.230 ppm 0.74 0.555fstab2 1.00000 3.00·10-3 50 32.9 0.0987 ppm 0.32 0.102nNofm 0.0368229

mole 17.7·10-6

Degrees of freedom

Corr.-coeff.

nN2 22.001383 mole

404·10-6 mole

mA2 1.044360 g 503·10-6 g MT2 27.793260

g/mol 553·10-6 g/mol

xA1 2.00358 % 3.82·10-3 % xN1 97.99582 % 2.71·10-3 % MA 17.030560

g/mol 128·10-6 g/mol

50 -1.93 -247·10-6 ppm

0.00 0.000

MN 28.0134800 g/mol

80.8·10-6 g/mol

50 1.17 95.0·10-6 ppm

0.00 0.000

mN2 616.3353 g 0.0112 g WbalA2 1.0443600 g 12.6·10-6 g 82 31.5 397·10-6

ppm 0.00 0.000

δWug 0.0 ug 50.0 ug 50 31.5·10-6 1.57·10-3 ppm

0.01 0.000

δWhadug 0.0 ug 500 ug 50 31.5·10-6 0.0157 ppm 0.05 0.003WbalN2 616.335310

g 133·10-6 g 74 -0.0533 -7.07·10-6

ppm 0.00 0.000

δWg 0.0 mg 10.0 mg 50 3.96·10-3 0.0396 ppm 0.13 0.016

δWhad 0.0 mg 5.00 mg 50 3.96·10-3 0.0198 ppm 0.06 0.004nA1 0.470178

mole 657·10-6

nT1 23.45989 mole

1.06·10-3 mole

fpurity 0.9997000 66.5·10-6 50 32.5 2.16·10-3 ppm

0.01 0.000

fhomo1 1.000000 900·10-6 50 32.5 0.0293 ppm 0.09 0.009fstab1 1.00000 1.00·10-3 50 32.5 0.0325 ppm 0.11 0.011nN1 22.989714

mole 405·10-6

mA1 8.0074 g 0.0112 g mN1 644.0219 g 0.0112 g

WbalA1 8.007400 g 205·10-6 g 91 4.06 832·10-6 ppm

0.00 0.000

WbalN1 644.021900000 g

230·10-9 g 98 0.0 0.0 ppm 0.0 0.0

CK46 32.913 ppm 0.309 ppm 124

Result: Quantity: CK46 Value: 32.91 ppm Expanded uncertainty: ±0.62 ppm Coverage factor: 2.0 Coverage probability: 95.45%

Annex D: Measurement report CERI

Report Form CCQM-K46 Ammonia in nitrogen Laboratory name: Chemicals Evaluation and Research Institute, Japan

(dd/mm/yy) Result (μmol/mol)

NH3 3/3/2007 32.679 0.069 3

NH3 4/3/2007 32.740 0.059 3

NH3 5/3/2007 32.907 0.076 3

NH3 5/3/2007 32.882 0.1132 3

NH3 6/3/2007 32.840 0.064 3

(μmol/mol) Expanded Uncertainty Coverage factor

NH3 32.81 0.38 2

Reference Method: Instruments for NH3 measurement Principles : NDIR (Type:CGT-7000, Make : Shimadzu corporation) Data collection : output of integrator of data Calibration Standards: Preparation : Gravimetric method Purity analysis ;

NH3, N2: The impurities in NH3 and N2 are determined by analyses and the amount of the major component is conventionally determined by,

iipure xX

where: xI = mole fraction of impurity i , determined by analysis N = number of impurities likely in the final mixture

Xpure = mole fraction ‘purity’ of the ‘pure’ parent gas Instrument Calibration: Table 1 concentration of PSMs

Concentration ( μmol/mol ) Component

R1 R1 R3

NH3 50.47 34.83 20.29

This procedure is for the determination of NH3 in a sample using NDIR. 1) Inject the calibration standard (R1) into NDIR. Record the output. 2) Inject the calibration standard (R2). Record the output. 2) Inject the sample to be tested in same manner as the calibration standard. Record the

output. 3) Inject the calibration standard (R3). Record the output. 4) Calculate the concentration of NH3. Following above procedure, 3 measurements are repeated subsequently in a day and

iterated for 5 days.

Uncertainty:

Uncertainty source

Estimate xI

Assumed distribution

Standard uncertainty u(xi)

Sensitivity coefficient cI

Contribution to standard uncertainty uI(y)

Repeatability of analysis 32.81 Normal (A) 0.185 1 0.185

Reference gas R1 preparation 50.47 normal (A) 0.025 1 0.025

Reference gas R2 preparation 34.83 Normal (A) 0.017 1 0.017

Reference gas R3 preparation 20.29 Normal (A) 0.010 1 0.010

total 0.188

Coverage factor: 2 Expanded uncertainty: 0.38 μmol/mol

Annex E: Measurement report METAS

Report Form CCQM-K46 Ammonia in nitrogen Laboratory name: METAS

Nominal composition: ammonia in nitrogen 30 to 40·10-6 mol·mol-1

(dd/mm/yy) Result (µmol·mol-1)

Combined standard uncertainty (µmol·mol-1)

Ammonia 05.03.07 32.30 0.19 100

Ammonia 06.03.07 32.49 0.19 100

Ammonia 07.03.07 32.04 0.19 100

Result Component Amount of substance fraction in the

test mixture (µmol·mol-1) Expanded uncertainty (µmol·mol-1)

Coverage factor3

Ammonia 32.27 0.36 2.0

Uncertainty budget The uncertainty budget has been calculated using GUM Workbench Pro software (version 2.3.2.36 beta).4 The main contributions to the combined standard uncertainty are:

• the standard uncertainty of the ammonia mass flow from the permeation unit with 42 %,

• the standard uncertainties of the purities of the two dilution gases with 26 %,

• the estimated standard uncertainties resulting from the interaction of the test mixture with the surface of the sampling line, consisting of the pressure regulator of the cylinder, the absolute pressure regulator and the sonic nozzle, with 21%

• and the standard uncertainties of the flow measurements with 11 %.

The standard uncertainty of the purity of the permeated ammonia contributes less than 0.1 % to the combined standard uncertainty.

2 If more than three measurements are taken, please copy and insert a table of the appropriate format as necessary 3 The coverage factor shall be based on approximately 95% confidence. 4 The complete budget is available on request

Also negligible are the standard deviations of the readings of the ammonia analyzer. With a 100 replicates the relative standard deviation of the mean of the readings is only 0.02 %.

Analysis Method A commercial photo-acoustic NH3-analyzer was calibrated with NH3 calibration standards in the range from 510 to 594 nmol·mol-1 NH3 in N2 for measurements 1 and 2 and with NH3 calibration standards in the range from 529 to 571 nmol·mol-1 NH3 in N2 for measurement 3.

The VSL-test mixture D751986 was dynamically diluted with nitrogen BIP Plus (Air Products) by a factor of about 0.017 such that the expected amount of substance fraction of the sample lies within the validated and calibrated range of the analyzer and of the METAS Primary micro-gravimetric Standard. The flow of the test mixture was set using a sonic nozzle. The upstream pressure at the sonic nozzle was kept constant by an absolute pressure regulator. The flow of the dilution gas was regulated with a mass flow controller and measured using a molbox-molbloc system. The resulting gas mixture was measured with the photo-acoustic NH3-analyzer and the amount of substance fraction calculated by linear interpolation in agreement with ISO 6143:2001(E).

Calibration Standards The calibration standards were produced by the METAS primary micro gravimetric standard and a NH3 permeation unit with purity ≥ 99.99 %. The total N2 dilution gas flow was measured by a molbox-molbloc system. The nominal N2 purity was ≥ 99.999 %. The ammonia mass flow of the permeation unit was approx. 550 ng·min-1 at 30 °C.

The two molbox-molbloc systems and the sonic nozzle were calibrated with the METAS Primary Standard for low gas flows.

Sample Handling An electro polished stainless steal pressure regulator with a flushing system was used for dispensing the VSL test mixture. Several flushing cycles with N2 and the NH3 test mixture were carried out.

After stable readings by the NH3 analyzer, data were sampled for at least 25 minutes.

Purity analysis of the dilution and permeation gases

Method An ion-molecule-reaction mass spectrometer was used for analysing the residual amount of substance fraction of ammonia in the used N2 qualities and to identify and quantify possible impurities in ammonia permeators. The ionising gases with the corresponding ionisation energies were: Hg (10.4 eV), Xe (12.13 eV) and Kr (14 eV) and Hg with high acceleration voltage (Ue high) for the unspecific ionisation of most analytes.

The calibration of the MS for ammonia in N2 was performed with a calibrated permeation unit with a mass flow of 100.4 ng·min-1 at 30 °C in a temperature controlled oven. The amount of substance fraction was set at three points in the lowest accessible range of 49 to 100 nmol·mol-1 by two calibrated mass flow controllers for the N2 carrier and dilution gas flows of the permeation oven, respectively. The MS parameters with Xe ionisation were optimized for maximum signal and minimal drift for the mass of 14NH3 at m/z=17.027 mu/e. The mass scale and mass resolution of the quadrupol MS were optimised with the signal from 14NH3 with the least contribution from the ubiquitous neighbouring H2

16O (m/z=18.011 mu/e). The conditioning times derived from the NH3 signal were at least 30 min.

Residual ammonia in dilution gases The limit of detection calculated from the intercept of measurement function at XNH3 = 0 and the 3 σ criteria from the measurement of the N2 BIP Plus (Air Products) with a specified purity ≥ 99.9999 % was 1.9 nmol·mol-

1 with typical sampling times of 6 min. The residual NH3 amount of substance fractions in the dilution gases were calculated with the B_LEAST program (Vs. 1.11, 1999, in agreement with ISO 6143:2001(E)) by linear regression and with the limit of detection being the least detectable amount. For both nitrogen gases N2 99.999 % and N2 BIP Plus the residual NH3 amounts were below the limit of detection.

Permeation gases Survey mass spectra with the three available ionisation energies and with Hg (Ue high) were recorded within a mass range of 10 to 195 mu/e. The sample was a NH3 permeation unit in the permeation oven with an N2 carrier flow of 300 ml·min-1. The model permeation unit used for these investigations had a permeation rate of about 997 ng·min-1 at 30 °C resulting in a NH3 amount of substance fraction of about 3070 nmol·mol-1. It was from the same producer and from the same production batch with identical geometry as the unit used for the calibration of the photo-acoustic NH3-analyzer. The high permeation rate was intentionally chosen to enhance the sensitivity of the method for detecting trace impurities. Before storing of the spectra the long conditioning times for NH3 were taken into account. Time scan spectra of the major suspected impurities of gas flows through the oven with and without the permeation unit and of the dilution gases alone were recorded.

Results In addition to NH3 H2O, O2 and CO2 were detected in the reference gas mixture from the oven. Because H2O, O2, and CO2 were about equally present in N2 after flowing only through PFA (Copolymer of Perfluoralkoxy and Tetrafluorethylene) tubes, it is concluded that the observed substances are not impurities from the permeation unit. The increase of the O2 amount of substance fraction in an N2 flow of 485 ml·min-1 in a model PFA tubing with comparable dimensions as the tubes in the oven was estimated to be 4 μmol·mol-1. With O2 having the highest permeation rate of the 3 substances through PFA, the amount of the other traces is supposed to be lower. Since NH3 exists in ambient air in combination with H2O, O2, CO2 they are unlikely to react with NH3 within the few seconds between the mixing and the detection. Therefore no significant loss of NH3 due to the detected trace amounts of impurities in the dilution gases from the used PFA tubes is expected.

Annex F: Measurement report VNIIM

RESEARCH DEPARTMENT FOR THE STATE MEASUREMENT STANDARDS IN THE FIELD OF PHYSICO-CHEMICAL MEASUREMENTS

Key Comparison CCQM-K46 Ammonia in Nitrogen REPORT Date: 26.06.07 Authors: L.A. Konopelko, Y.A. Kustikov,V.V. Pankratov, D.N. Selukov,V.A. Petrov, E.V. Gromova

Reference method: UV absorption Instrument: Spectrophotometer Lambda 900 (“Perkin Elmer”, USA) Length of the cell – 14 cm. Calibration standards Characteristics of pure substances used for preparation of the calibration standards are

shown in table 1. Table 1 – Description of pure components

Component Mole fraction 10-6 mol/mol

Standard uncertainty 10-6 mol/mol

NH3 999900 60 N2 999990,0 0,5

All standard gas mixtures were prepared in aluminium cylinders with Aculife IV

treatment, V= 5 L.

Weighing was performed on the balances 81-V-HCE-20kg (hnu-Voland, USA). Experimental standard deviation for 5 L cylinders: 8 mg.

Preparation of standard gas mixtures was carried out in 3 stages 1 stage: Preparation of the first stage gas pre-mixtures NH3/N2 with ammonia mole fraction on

the level of 4 %. Verification of mole fraction was carried out by UV absorption gas analyzer “OAC-

3600” (“Monitoring Ltd”, Russia). Standard deviation for each measurement series was not more than 0,1 %.

2 stage: Preparation of the second stage gas pre-mixtures NH3/N2 with ammonia mole fraction on

the level of 0,2 %. Verification of mole fraction was carried out by UV absorption gas analyzer “OAG”

(“Monitoring Ltd”, Russia). Standard deviation for each measurement series was not more than 0,1 %.

3 stage: Preparation of standard gas mixtures NH3/N2 with ammonia mole fraction of 32-34

ppm. There were prepared 4 standard gas mixtures.

Verification of mole fraction was carried out by UV absorption analyzer “Lambda” (“Perkin Elmer”, USA). Standard deviation for each measurement series was not more than 0,2 %.

The characteristics of calibration standards are shown in table 2. Table 2 – Characteristics of calibration standards

Standard gas mixture N

Component Assigned value, 10-6 mol/mol

Standard uncertainty, 10-6 mol/mol

NH3 32,73 0,18 1 N2 balance - NH3 33,23 0,18 2 N2 balance - NH3 34,62 0,18 3 N2 balance - NH3 33,15 0,18 4 N2 balance -

Instrument calibration Linear regression by 4 calibration points (4 standard gas mixtures with similar

concentrations) was used for instrument calibration. There were made 4 independent measurements under repeatability conditions with 4

independent calibrations. One single measurement consisted of 4 sub-measurements.

Sample handling Prior to measurements the cylinder was stabilized to room temperature. Results of measurements Results of measurements of ammonia mole fraction in cylinder № D7521114 are shown

in the table 3 Table 3 - Results of measurements of ammonia mole fraction in cylinder № D752114

(dd/mm/yy) Result (10-6 mol/mol)

NH3 21/05/07 33,72 0,3 4

NH3 25/05/07 33,65 0,3 4

NH3 31/05/07 33,65 0,3 4

NH3 08/06/07 33,68 0,3 4

Evaluation of uncertainty of measurements Total standard uncertainty of ammonia mole fraction was calculated on the base of the

following constituents: - total standard uncertainty of ammonia mole fraction in standard gas mixture

(including uncertainty of weighing of parent gases and pre-mixtures, uncertainty in the purity of the parent gases);

- standard deviation of linear regression; - standard deviation of the measurement result of ammonia mole fraction in

investigated gas mixture in cylinder № D752114 Uncertainty budget for ammonia mole fraction in gas mixture in the cylinder №

D752114 is shown in the table 4.

Table 4– Uncertainty budget for ammonia mole fraction in gas mixture in cylinder № D752114

№ Source of uncertainty Type of evaluation

Standard uncertainty, % relative

Preparation of the first pre-mixtures A 0,055

Preparation of the second pre-mixtures A 0,044

Preparation of the final mixtures A 0,55

Impurities in N2 A;B 0,000045

1 Preparation of standard gas mixtures

Impurities in NH3 A;B 0,0058

2 Standard uncertainty of calibration A 0,4 3 Standard deviation of the measurement result A 0,3 Combined standard uncertainty 0,75 Expanded uncertainty 1,5

Final result of measurements Final result of measurements of ammonia mole fraction in investigated gas mixture is

shown in the table 5

Table 5

Component Result (10-6 mol/mol)

Expanded Uncertainty

(10-6 mol/mol)

Relative Expanded

Uncertainty (%) Coverage factor

NH3 33,7 0,5 1,5 2

Annex G: Measurement report VSL

Report Form CCQM-K46 Ammonia in nitrogen Laboratory name: VSL

Standard deviation (μmol/mol)

Ammonia 2007-04-25 33,77 0,30 5

Ammonia 2007-06-01 33,80 0,34 2 x 5

Ammonia 2007-06-06 33,84 0,35 2 x 5

(μmol/mol) Expanded Uncertainty

(μmol/mol) Coverage factor

Ammonia 33,8 0,7 2

Reference Method: The CCQM-K46 mixture is compared to gravimetric standards using an Innova 1312 Multigas Analyzer based on Photo Acoustic Spectroscopy (PAS).

Instrument Calibration: The following VSL PSMs were used to calibrate the PAS analyser. All PSMs contained NH3 in a matrix of nitrogen, similar to the sample mixture.

Cylinder No Gravimetric composition (μmol/mol)

standard uncertainty (μmol/mol)

VSL118876 29,99 0,18 VSL328516 30,01 0,18 VSL328518 40,00 0,24 VSL118880 40,04 0,24 VSL328515 60,03 0,30 VSL238433 79,19 0,40

Cylinder No Gravimetric composition (μmol/mol)

standard uncertainty (μmol/mol)

VSL206351 80,13 0,40 VSL309544 100,01 0,45 VSL328438 200,09 0,60 VSL118863 200,72 0,60 VSL118862 299,10 0,75 VSL206346 301,18 0,75

Sample handling: Each cylinder was equipped with a stainless steel pressure regulator that was adequately purged. A flow of approx. 1,0 L/min was flushed for fifteen minutes, through FEP tubing, to the PAS analyser before the readings were taken.

Calibration Standards: The PSMs used for calibration are prepared from pre-mixtures in accordance with ISO 6142: 2001 (Gas analysis - Preparation of calibration gas mixtures - Gravimetric method). After preparation the composition was verified. The concentration of the PSMs has been validated with a continuous syringe injection technique according to ISO 6145:2004 (Gas analysis - Preparation of calibration gas mixtures using dynamic volumetric methods - Part 4: Continuous syringe injection method).

Evaluation of measurement uncertainty: The listed gravimetric uncertainty is a combined standard uncertainty and includes:

The uncertainty in the weighings The uncertainty on the purity analysis Stability issues; the dominant source of uncertainty in these type of mixtures

The listed standard deviations in the three measurements come from Generalized Distance Regression (GDR), already taking into account the uncertainties of the PSMs and the standard deviation in the responses. The expanded uncertainty in the result is calculated by taking the square root of the average variance of the individual measurements, multiplied with a coverage factor of 2.

Annex H: Measurement report NIST U.S. Department of Commerce

National Institute of Standards and Technology Chemical Science and Technology Laboratory

Analytical Chemistry Division Gaithersburg, MD 20899

REPORT OF ANALYSIS

May 18, 2007

Analysis of One Compressed Gas Mixture for CCQM-K46, Ammonia in Nitrogen

Submitted to:

Stephen A. Wise, Chief Analytical Chemistry Division

The Gas Analysis Working Group of the CCQM initiated a Key Comparison (CCQM-K46) on the analysis of ammonia in nitrogen. VSL, the Dutch Metrology Institute, was selected as the coordinating laboratory, which produced the gas mixtures, prepared the protocol, and conducted the comparison. VSL produced a batch of cylinders containing nominally the same concentration of ammonia. One each of these gas mixtures, contained in an aluminum cylinder, was sent to each participant for analysis. The gas mixture was to be between 30 µmol/mol and 40 µmol/mol ammonia in a balance of nitrogen. NIST analyzed this gas mixture, using a system utilizing permeation tubes, as the primary reference, and then returned the cylinder to VSL for reanalysis. Permeation System A permeation system was used to produce dynamic primary reference gas mixtures to calibrate the instrument used in this comparison. The permeation system consisted of a magnetic suspension microbalance (NIST #624294), a water bath (NIST #622824) to control the temperature of the permeation environment, a pressure controller to maintain a stable pressure in the permeation environment, and molbloc flow elements to measure the flow from mass flow meters. The microbalance had a resolution of 1 µg, with a repeatability of approximately 3 µg. This microbalance is capable of engaging and disengaging the load (the permeation tubes) through a magnetic suspension mechanism. This allows the balance to be recalibrated and the zero read at any time during the experiment. For this work the balance was recalibrated, and the zero determined, every 30 minutes. The mass of the permeation tubes were determined prior to each recalibration. The reading from the balance was averaged until 10 consecutive points resulted in a standard deviation of less than 6 µg.

The permeation environment was water jacketed and over wrapped with insulation to stabilize temperature in the permeation space. The water bath set point was 40 °C, and the internal environment temperature was determined to be (39.56 ± 0.02) °C. Long term drift of the temperature was measured to be less than 0.08 °C. The pressure in the permeation environment was also pressure stabilized, in order to eliminate buoyancy effects. Two controlled flows, both nitrogen, were directed into the permeation system. The first flow of 500 sccm was directed to the top of the permeation environment, and flowed over the permeation tubes. The second flow was varied to produce the differing concentrations required to produce the calibration curve. This flow varied between 1000 sccm and 1300 sccm, depending on the concentration of ammonia required. The total flow into the permeation system (Flow 1 plus Flow 2) was measured by a calibrated 5000 sccm molbloc. This molbloc was calibrated using a piston prover primary calibrator and gravimetric calibration. The nitrogen gas was supplied from boil off from a liquid nitrogen container. Permeation Devices A total of 4 permeation tubes, each 10 cm long and rated at 3100 ng/min ammonia at 30 °C, were used. These tubes were inserted into the permeation environment and allowed to permeate for one week at 25 °C prior to use. An identical permeation tube, purchased at the same time and from the same lot was used to determine the purity of the ammonia permeating from the device. A RGA-200 was used to determine the purity of the ammonia. The permeation tube was inserted into a glass container which was subsequently evacuated to < 0.05 torr. The tube was kept under this vacuum for 1 week in order to eliminate ambient water, and to remove the natural nitrogen and oxygen in the permeation tube which is in normal equilibrium with the atmosphere. After a one week purge, the following mass spectra was obtained.

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64-8-1.8x10

-88.0x10

-71.8x10

-72.8x10

-73.7x10

-74.7x10

-75.7x10

-76.7x10

-77.7x10

-78.6x10

-79.6x10

TorrAnalog Scan

<LIN> 2/12

X = 65.0 Y = 1.90e-010

This mass spectra shows the primary peaks associated with ammonia along with peaks associated with hydrogen and nitrogen. It is believed that the ammonia was disassociated within the RGA into hydrogen and nitrogen, and that these peaks do not indicate impurities. The purity of the ammonia gas was determined to be 99.4 %, with 0.1 % argon, 0.1 % oxygen, and 0.4 % carbon dioxide. Since these contaminants would not result from permeation, they were determined to be artifacts, perhaps from some leaks in the system, and the purity was taken at 100 %. However, since there is some uncertainty associated with this determination, a standard uncertainty of 0.35 % was assigned (0.6 %, rectangular distribution). FTIR was also used to assess the purity of the ammonia by directing the flow from the permeation system through a 10 m cell, and scanning for 4 hours. This spectra, shown below, shows some water contamination which was not above the background levels. There was no evidence of carbon dioxide contamination.

Calibration From a quick analysis of the comparison cylinder, it was determined that the concentration of the mixture was at the lower range of the 30 µmol/mol to 40 µmol/mol spread. Therefore, the permeation system was setup to provide mixture concentrations of nominally 28 µmol/mol, 30 µmol/mol, 32 µmol/mol, 34 µmol/mol and 36 µmol/mol ammonia in nitrogen. The uncertainty of the gas mixture concentrations from the permeation system were assessed to be 0.2 % relative. This uncertainty is dominated by the flow calibration, and flow stability uncertainty. A total of three analytical runs were conducted over three days. Each analytical run consisted of analyzing the calibration gas mixtures a minimum of three times, and sampling the comparison cylinder before and after each calibration series. The uncertainty associated with the reproducibility of the analytical measurement was approximately 0.05 % relative. The data from each analytical run was processed using the GENLINE algorithm. Results The results of the three independent analytical runs are given in Table 1. The uncertainty in this table are results only from the GENLINE algorithm, and thus are from the instrument reproducibility and calibration gas mixture uncertainty only.

Table 1; Analytical Results Analytical Run

Value (µmol/mol)

Uncertainty (k=1)

1 32.46 0.02 2 32.43 0.02 3 32.22 0.02 Average 32.37

The uncertainty of this analysis is dominated by the purity uncertainty of 0.35 % relative. The standard error of the mean of the three values is 0.08 µmol/mol. Combining the three sources, where k=2, yields a final expanded uncertainty of 0.28 µmol/mol (0.9 % relative). The final submitted value for CCQM-K46, cylinder D751992, is (32.37 ± 0.28) µmol/mol. Prepared by: Franklin R. Guenther, Ph.D. Supervisory Research Chemist Reviewed by: George C. Rhoderick Research Chemist

An additional study of NH3 gas standards

Maruyama masaaki, Uehara shinji, CERI

1. Introduction

CERI’s result was approximately 5% lower than preparation value on CCQM-K46 (NH3 in

N2). Each participant used Scott cylinders or Luxfer cylinders or Permeation system. CERI

used Luxfer cylinders. Therefore we planned an additional study to check the difference

between treatment by Luxfer and Scott.

We had already verified that NH3 concentration was stable in passivated Luxfer cylinders

treated by Luxfer. (The stability in 6 month is less than 1%.) We need to show the stability of

NH3 gas standards in Scott cylinders treated by Scott. But Scott cylinders are not allowed to

fill gases because of Japanese regulation. So, we used Luxfer cylinders treated by Air Liquide

instead of Scott cylinders treated by Scott in this study. Treatment by Air Liquide is the same

as that by Scott. These cylinders aren’t commercial. But we could get them in collaboration

with K. K. Air Liquide Laboratories and Air Liquide America Specialty Gases LLC. (Scott

Specialty Gases is now Air Liquide America Specialty Gases LLC.)

2. Experiment

2.1 Stability on non-passivated Luxfer cylinders treated by Air Liquide

(1) Cylinders were filled with N2 after treatment.

(2) Evacuated the cylinders (below 5 Pa).

(3) Transferred NH3 gas standard into the cylinder.

Parent gas was one single approximately 35μmol/mol. (4) Calibrated the samples by PSMs

Right after transfer, 3 days later and 3 weeks later

2.2 Stability on passivated Luxfer cylinders treated by Air Liquide

(1) Passivation

Experiment 2.2 was done after the Experiment 2.1

(2) Evacuated the cylinders (below 5 Pa).

(3) Prepared 35μmol/mol NH3 gas standards (7MPa) into the cylinders by gravimetric.

Parent gas was one single 1500μmol/mol

(4) Calibrated samples by PSMs.

2 days later, a week later and 3 weeks later

3. Results

3.1 Non-passivated cylinders

Table 1 shows the result of Experiment 2.1

All of the calibrated values were lower than the concentration of parent gas

The calibrated values of ‘3 days later’ and 3 weeks later’ were close.

Table 1 Results of non-passivated cylinders

Calibrated values (μmol/mol)

(Relative deviation(%))

Cylinder

Numbers

Parent

(μmol/mol) Right after transfer 3 days later 3 weeks later

CPB-18498 34.85 32.04

(-8.77)

(-12.78)

(-13.59)

CPB-18499 34.85 30.28

(-15.09)

(-21.22)

(-22.67)

3.2 Passivated cylinders

Table 1 shows the results of Experiment 2.2

All of the calibrated values were almost consist with gravimetric value.

Table 2 Results of passivated cylinders

Gravimetric Calibrated values (μmol/mol)

(Relative deviation(%)) Cylinder

Numbers

(μmol/mol) 2 days later A week later 3 weeks later

CPB-18498 35.26 35.08

(-0.51)

(-0.89)

(-0.77)

CPB-18499 34.89 34.70

(-0.55)

(-1.04)

(-0.87)

4. Discussion

4.1 Non-passivated cylinders

NH3 concentration was decreased even when ‘the right after transfer’.

But after ‘3 days later’, concentration was considered to become stable since calibrated values

of ‘3 days later’ and ‘3 weeks later’ were close.

From those results, on cylinders treated by Air Liquide without passivation, NH3

concentration was considered to be decreased within 3 days after filling NH3 gas standards.

4.2 Passivated cylinders

NH3 concentration was considered to be stable since all of the calibrated values were almost

consistent with gravimetric values.

5. Conclusion

If the concentration of passivation is more appropriate, prepared concentration will be more

stable in Luxfer cylinder treated by Air Liquide.

These results are consistent with the result of Luxfer cylinder treated by Luxfer.

Consequently, these results suggest that there are few differences about stability between

treatment by Luxfer and Scott.

CCQM K46 – Ammonia in nitrogen

Overview of the results and report of additional actions carried out at VSL

Gerard Nieuwenkamp, Rob Wessel

VSL B.V., Department of Chemistry, Thijsseweg 11, 2629 JA Delft, the Netherlands

Report number S-CH.09.07

Delft, 14 April 2009

Table of contents

1 Introduction 1

2 Validation prior to shipment 1

2.1 Validation with Primary Standard gas Mixtures in cylinders 1

2.2 Validation by use of a continuous syringe injection method 2

2.3 Validation of the continuous syringe injection method 4

3 Results of participating NMI’s 5

4 Discussion: 6

4.1 Permeation 6

4.2 Calibration standards in cylinders: Doping vs not doping 6

5 Conclusions 7

1 Introduction

The results of the CCQM K-46 have not led to a clear consensus between the participants. This document summarises the analytical results so far, including the validation of the cylinders prior to shipment and the additional actions that have been performed by some of the participants. This overview may be of help in the discussion how to proceed with the results of K-46.

2 Validation prior to shipment

A total number of 7 mixtures with a nominal amount of substance fraction of 34 · 10-6

mol/mol NH3 in N2 has been prepared gravimetrically by VSL (ISO6142). These mixtures have been analyzed against a set of VSL PSMs and in addition by use of continuous syringe injection (ISO6145-part 4).

2.1 Validation with Primary Standard gas Mixtures in cylinders

The new mixtures for K-46 have been prepared in October 2006. The verification is performed by a bracketing method using two PSMs of 30 · 10

-6 mol/mol NH3 in N2 and two PSMs of 40 · 10

-6 mol/mol

NH3 in N2. The PSMs used in this measurement have been prepared in 2005 and 2006. VSL has an annual maintenance schedule, in which new mixtures are prepared every year and compared to the existing standards. No significant issues of long-term instability have been observed.

Table 1: verification results for the K-46 mixtures in comparison to Primary Standard gas Mixtures

cylinder number

gravimetry µmol/mol

October analysis

November analysis

December analysis

January analysis

D751992 33.929 33.9 33.8 33.9 33.6

D752114 33.983 34.2 34.2 34.2 34.1

D751916 34.028 34.2 34.1 34.0 34.0

D751986 34.057 34.0 33.8 33.6 33.6

D751946 34.091 34.0 33.8 33.7 33.9

D751929 34.104 34.2 34.1 33.7 34.2

D751936 34.113 34.0 34.1 33.6 34.0

Figure 1: verification results for the K-46 mixtures in comparison to Primary Standard gas Mixtures, expressed as recovery related to the gravimetric value

Conclusions:

- The analytical results for the new mixtures for K-46 overlap within the uncertainty with their gravimetric values.

- By using older PSMs, this also proofs that there is no significant sign of long term instability of this type of mixtures in this type of cylinders.

- However, these results do not yet prove that there is no initial loss in the preparation of this type of mixtures.

2.2 Validation by use of a continuous syringe injection method

In order to detect the presence of systematic errors (e.g. initial loss caused by absorption) in the gravimetrically prepared mixtures, the new set of mixtures has been verified by calibration with an independent method. A gas-tight syringe is initially filled with pure NH3 and injected in a flow of nitrogen. The flow is adjusted to make a dilution with a nominal concentration of 300 · 10

-6 mol/mol

NH3 in N2. The flow is measured by using a calibrated mercury piston sealed flow meter. The monitor response of the dynamically prepared mixture is compared to the response of two gravimetrically prepared cylinders of comparable concentration. Figure 2 shows these results, expressed as monitor response / ppm NH3.

mean value validation analyses

0 2 4 6 8

cylinder number

ecovery

Figure 2: Comparison of syringe injection with gravimetrically prepared 300 · 10-6 mol/mol NH3 in N2

mixtures

After the comparison at the 300 · 10-6

mol/mol NH3 in N2 level, the syringe is filled with a gravimetrically prepared 5 · 10

-2 mol/mol NH3 in N2 mixture. The flow is adjusted to make a 30 · 10

mol/mol NH3 in N2 mixture by dilution. The dynamic mixture is compared to two gravimetric mixtures; one of them prepared in 2005, and the other already in 2001.

Figure 3: Comparison of syringe injection with gravimetrically prepared 30 · 10-6

mol/mol NH3 in N2 mixtures

Conclusions:

- The results of the dynamically prepared mixtures overlap with the gravimetric values, both at the 300 · 10

-6 mol/mol NH3 in N2 and the 30 · 10

-6 mol/mol NH3 in N2 level.

- This confirms the overall consistency of the different NH3 concentrations (pure, 5%, 300 ppm, 30 ppm) that have been used in this comparison.

- However, there may have been an unidentified systematic error in the performance of the dynamic method.

0 2 4 6 8 10 12

tive r

syringe

gravimetry '01

gravimetry '05

0 2 4 6 8 10 12 14

syringe

gravimetry '01

gravimetry '05

2.3 Validation of the continuous syringe injection method

To validate the method, all individual parts of the method can be calibrated, but that is still no 100% guarantee that the system works as expected. To overcome any doubt, the continuous syringe injection method has been validated as a whole.

The syringe has been filled with a gravimetrically prepared 5 · 10-2

mol/mol CO in N2 mixture. The flow is adjusted to generate dilutions of respectively 80 and 40 · 10

-6 mol/mol CO in N2 mixture. These

dynamically generated mixtures have been compared to PSMs of CO in N2 of similar concentration.

mol/mol CO in N2 mixtures

Conclusions:

- The results for the dynamically generated CO mixtures overlap with the gravimetric value of CO mixtures in cylinders (these values are beyond any doubt, based on previous Key comparisons)

- The dynamic generation of gas mixtures by this continuous syringe method appears to be OK, with an expanded uncertainty of 1% relative.

- The combined set of validation results (2.1, 2.2. and 2.3) gave enough evidence and confidence to proceed with the shipment of the cylinders to the participating laboratories.

0 2 4 6 8 10 12

tive r

syringe

gravimetry

0 2 4 6 8 10 12

tive r

syringe

gravimetry

3 Results of participating NMI’s

Figure 6 shows the result of the key-comparison.

Figure 6: Results of participants in CCQM-K64 Ammonia in nitrogen

The participating institutes used different methods of calibration. Roughly the methods can be divided in three groups:

- Calibration with NH3 mixtures in cylinders

o VNIIM, NMi and NPL

- Calibration with NH3 mixtures in “doped” cylinders

o CERI and KRISS

- Calibration with permeation standards

o NIST and Metas

The results of the participants using the same type of calibration are in good agreement with each other.

For the discussion about the different results with the different calibration methods it may help to give an overview with the obvious advantages vs. disadvantages of the selected methods.

CCQM-K46 NH3 in N2

NIST VNIIM NMi Metas CERI NPL KRISS

relative d

eviation (%)

Table 2: Advantages and disadvantages of the different calibration methods

Calibration by: Advantage Disadvantage

NH3 mixtures in cylinders

Sample and calibration gas are sampled under the same conditions

Possible absorption to the cylinder surface

NH3 mixtures in pre-filled cylinders

Minimized possibility for absorption Possible increased concentration by desorption of NH3 from the surface

Permeation

No stability/absorption issues related to cylinder surface

Use of a well conditioned calibration compared to a less conditioned sample

4 Discussion:

4.1 Permeation

- After the presentation of the K46 results, NIST stated that the cylinder has a side connection (DIN-1) that is different than their usual size. This means that the pressure regulator used for this comparison was not of the preferred type. VSL has sent one of their regulators to NIST. With this pressure regulator NIST had a smaller discrepancy (-3% rel) for the K46 mixture.

- Metas send a report of their permeation set-up. They did not find any source in their measurements that could cause the observed discrepancy between the K46 mixture and their calibration

From our personal point-of-view there is no reason to question the quality of the permeation system as used by the two institutes. The dynamic generation system will provide a constant and well-defined calibration gas in a good conditioned sampling system. However, the gas sample from a relatively small cylinder is taken under significant different conditions. A (non-conditioned) pressure regulator, new tubing and a mass flow controller are required to take a representative sample from the cylinder. Due to the limited amount of gas in the cylinder, the flushing time is limited. It seems possible that the discrepancy is caused by the comparison of two different systems (comparing apples to pears).

4.2 Calibration standards in cylinders: Doping vs not doping

One of the uncertainty sources in the preparation of reactive gas mixtures by gravimetry is the initial loss caused by absorption of component to the cylinder surface. Much work is performed by the commercial gas suppliers to treat the internal cylinder surface in order to avoid absorption. Treatments with the names of Aculife, Brillanté, Spectra Seal, Quantum, AlphaTech, etc., are available. VSL has worked in many co-operation studies with the suppliers to find a suitable treatment for every component in the scope of VSL‘s activities. Apart from that, experiments have been performed by doping the cylinder with an elevated concentration of the component prior to the gravimetrically preparation of a mixture (e.g. NO2, ethanol). Also doping with other components has been tried (doping with SO2 for NO2 mixtures and vice versa).

From a principal point-of-view VSL chooses not to use doped cylinders for their reference materials. Mixtures are prepared very accurate with gravimetry, so “what goes in” is very well-defined. By selecting an appropriate cylinder treatment this should be equal to “what comes out”. This is demonstrated by comparison with independent dynamic methods, or by comparison to mixtures from sister NMI’s.

In case that significant absorption is expected to take place, the cylinders can be doped to occupy active sites at the surface. However, this opens the possibility of desorption, because of changing equilibrium conditions in the cylinder. This can be very complex to integrate in the uncertainty budget of the mixture. In relation to a changing pressure during the gas mixture‘s lifetime the desorption behaviour becomes even more complex.

For VSL, the annual repetition of mixture preparation gives confidence in the absence of significant absorption, considering that the amount of absorption will be cylinder dependant and therefore random spread over a certain interval. Discrepancies as found in K46 between the participants have not been observed in the results of VSL mixtures.

VSL performed an experiment with doping of NH3 cylinders. Mixtures prepared in doped cylinders have been compared to mixtures in "normal" cylinders. At the 35 ppm level, a difference of 2% relative was observed in the results, which is in the same order of magnitude as the uncetainty of this comparison. So, this experiment does not confirm the results of KRISS and CERI being at approx. -4%, but it also doesn't confirm that no absorption at the cylinder surface takes place. In addition, an ordinary VSL cylinder filled with 300 · 10

-6 mol/mol NH3 in N2 has been evacuated and

filled with pure nitrogen. After two weeks the NH3 concentration is analysed by CRDS to be less than 10 · 10

-9 mol/mol. This means that no desorption has taken place. Does this also proof that no

absorption has taken place in the initial cylinder?

In an additional report of CERI, the results of using doped cylinders are described. From the new mixtures that are prepared every 6-months, they obtain the same kind of confidence in their values as was described for VSL before. The presence of a significant effect of desorption is even unlikely as was the absorption in the VSL figures. So both methods look solid.

CERI also reported additional research on a commercially passivated cylinder similar to that used by VSL, NPL and VNIIM. Using this passivated cylinder without doping caused a drop in concentration of more the 20%. This is not comparable to the differences seen in this key comparison.

5 Conclusions

The spread in results as found in K46 indicate that the 3 different methods used give similar results for similar methods but do not overlap with each others within the given uncertainties. It is clear that all three methods have advantages as well as disadvantages were the size of the effect is summarized in table 3.

Table 3: Advantages and disadvantages and size of effect of the different calibration methods

Calibration by:

Advantage Disadvantage Size of effect

NH3 mixtures in cylinders

Sample and calibration gas are sampled under the same

conditions

Possible absorption to the cylinder surface

NH3 mixtures in pre-filled cylinders

Minimized possibility for absorption

Possible increased concentration by

desorption of NH3 from the surface

Permeation

No stability/absorption issues related to cylinder

surface

Use of a well conditioned calibration

compared to a less conditioned sample

International Comparison CCQM K46 – Ammonia in Nitrogen · 2010-08-23 · International...

Documents

Sintesis Ammonia

Trắc nghiệm đầu tư nước ngoài - ftu k46 bookbooming

Ammonia Recovery Simulation

International Comparison CCQM K23b – Natural gas …...International Comparison CCQM K23b – Natural gas type II Adriaan M.H. van der Veen1, Hima Chander1,Paul R. Ziel1, Ed W.B

Ammonia P-IV

International Comparison CCQM K23b – Natural gas type II

Ammonia Oxidizer Enrichment

K46 Nguyen Cam Tu Thesis

Uhde ammonia

User Manual ASUS K46-K56

Cara Menganalisa AMMONIA

FINAL Report Ammonia

Final Report CCQM-120 - BIPM€¦ · International comparison CCQM-K120.a Carbon dioxide in Air at background level (380-480) µmol/mol CCQM-K120.b Carbon dioxide in Air at urban

K46 2005 12

Ammonia MSDS Romana

171781496 Sintesa Ammonia

CCQM-K115 hCP Final Report - BIPM. Josephs Final Report CCQM‐K115 18.01.2017 1 ... * Bureau International des Poids et Mesures (BIPM) Sèvres, ... Final date to register October

CCQM-K90, Formaldehyde in nitrogen, 2 μmol mol Final report · CCQM-K90 Final Report v3.4.docx 26 Sep. 17 Page 4 of 67 . 2 Purpose . The CCQM-K90 comparison designed to evaluate

Russia Ammonia Market

SCES2340 P4 Ammonia Synthesis